AI Resource Concerns - Part 2
There are several real concerns when it comes to AI. One of them is about resources, specifically water and power. Datacenters or “server farms” require fresh water to cool down and power to work.
For years, data center energy requirements kept pace with new efficiency gains so there weren’t any dramatic changes. AI broke that pattern. Energy consumption in “accelerated servers” which are mostly AI, is projected to grow at 30% annually - versus 9% for non-AI servers.
One of the reasons we should worry about this is because the data center energy costs appear to be being passed on to the individual consumer. Of the electricity markets that recorded price increases, more than 70% of them are located within 50 miles of big data center activity. Amazon commissioned a study and proved that they are paying for all the electricity they consume. But the cost increases near data centers come primarily from grid infrastructure upgrades that are required to deliver that electricity. Those costs are billed differently and are often distributed across all the electricity users. So, yes, the data centers are paying their own electricity bills. But also yes, your bill may go up because of the data centers.
As more and more individuals and communities understand what’s at stake, they’re pushing back. Florida has new rate structures just for data centers, Oklahoma is proposing a moratorium on datacenters until 2029, Oregon is passing legislation requiring regulators to separate data center costs from other customer’s bills, and Ohio’s governor just paused all new data center tax breaks while lawmakers review the economic and infrastructure impacts. Ohio had projected the exemption would cost $136 million in 2025 and it ended up being 1.6 BILLION. And there’s a good chance that there will be a data center ban on the November Ohio ballot. In my town of Grove City, we just voted last week on a one-year moratorium on data centers until we could take a better look at what was being proposed.
These data centers are mostly unmanned so once the centers are up and running and the construction is done, they’re not generating any jobs. They’re loud – with thousands of fans running 24/7.
And then there’s the water usage. Data centers consume water using liquid cooling systems, and large data centers can use up to 5 million gallons per day. Microsoft consumed 1.69 BILLION gallons of water in 2022, representing a 34% jump over the prior year, with generative AI workloads believed to be at least partially responsible.
With recent news stories about US aquifers drying up, this quantity of water usage is seriously concerning.
There are other ways to cool the data centers and those are being explored as quickly as possible. But given the push to build build build to keep up with prospective demand, not many of those are going to be implemented in time.
In March of this year, several big tech companies (Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI) signed a pledge to pay to either build their own power infrastructure or to upgrade the existing grid infrastructure, use separate rate structures, and help prevent blackouts during emergencies. There were no commitments made around water consumption or the health impacts on surrounding communities, air quality or any sort of community input process. The pledge also has no binding enforcement. It’s entirely voluntary. There is no auditing, no penalties for non-compliance. It’s a PR commitment. And it’s being made while simultaneously racing to build faster than any of those commitments can realistically keep up with.
According to the memes I’m seeing on social media, using ChatBots and generating silly images are killing our planet. Eh. While asking ChatGPT for feedback on something uses ten times more energy than a standard google search, that number is still fractions of fractions. The individual user, even the world of individual users, are still a drop in the bucket compared to large corporations.
And it’s not like not using AI to answer a question is reducing your energy usage to zero. You’re STILL using energy asking a traditional search engine a question. Also keep in mind that you might need to do a lot of Google searches to find out information that an AI assistant consolidates into a single response.
A human assistant doing comparable work – researching, writing, problem solving - has their own footprint commuting, office heating and cooling, a computer running all day, lighting, a building's shared infrastructure. None of that gets talked about in the "AI uses water" conversation, because we simply don't think about the resource cost of human labor the same way.
There's also a time dimension that matters. A business owner outsourcing to a human assistant might get a response in hours or days. The iteration cycle is slow. With AI, they can draft a client proposal, get feedback on it, revise it, and finalize it in 20 minutes. That compression of time has real business value - fewer rushed late nights, faster turnaround on client communications, more headspace for the actual human work.
The primary AI data center usage is enterprise automation running millions of API calls per day, model training runs that cost tens of gigawatt-hours each, and infrastructure buildout driven by corporate competition. A significant share goes to AI embedded invisibly in advertising systems and social media recommendation engines. And yes, video games, entertainment, and content recommendation algorithms are real consumers too. The resource consumption isn't being driven by people asking chatbots questions. It's being driven by the largest corporations on Earth automating their operations at massive scale, and by tech companies racing to build the next generation of AI infrastructure. The individual user is a rounding error in that picture.
Listen, it’s not all about optimizing Google searches and creating targeted advertising. 2025 saw significant AI-driven scientific breakthroughs. From faster Alzheimer's diagnosis and detection to Google's AlphaGenome model for understanding diseases and drug discovery. AI is being used for climate modeling, power grid optimization, and weather forecasting at scales humans couldn't manage manually. Companies used AI-driven platforms to predict drug effectiveness, analyze complex genomic data, and shorten clinical trial timelines, and AI-enabled platforms analyze real-time patient data from wearables to personalize treatment plans. The potential in AI is… mind-bending.
And yes, the demand chain is real. Every person using ChatGPT, MidJourney, and Claude – that aggregated demand is what justifies the $364 billion investment in data center construction. The corporations aren’t building all of this in a void. They’re building it because hundreds of millions of people are using these tools daily and the usage keeps growing.
But there's an important distinction between being part of the demand and being responsible for how that demand gets met. You stream Netflix - you don't decide whether their data centers run on renewable energy or sit on top of an aquifer in Arizona.
A huge share of the data center energy goes into model training runs – which happens at the corporate level with zero input from the individual users. We have almost NO leverage over how efficiently these models are built, trained, or cooled. The corporations decide whether to use renewable energy, build efficient cooling systems, choose water-stressed locations for their data centers. Those decisions are being made in boardrooms and THAT’S where the pressure should be applied.
The corporate AI buildout is driven by competition. Microsoft, Google, Meta… they’re building infrastructure not for today’s demand but to WIN the AI race. And they’ll trample anyone in their way. This is why communities are standing up and demanding accountability and asking them to slow down.
While there is obviously a valid concern with the use of resources, I’d like to suggest we point the finger in the right direction. And the right direction is not at the individual using Claude to help write an email that uses a little more tact.
References
International Energy Agency. (2025). Energy and AI. IEA. https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai
International Energy Agency. (2026, April). Key Questions on Energy and AI.https://www.iea.org/reports/key-questions-on-energy-and-ai/executive-summary
Pew Research Center. (2025, October 24). What we know about energy use at US data centers amid the AI boom.https://www.pewresearch.org/short-reads/2025/10/24/what-we-know-about-energy-use-at-us-data-centers-amid-the-ai-boom/
Data Center Dynamics. (2023, September 12). Microsoft's water consumption jumps 34 percent amid AI boom.https://www.datacenterdynamics.com/en/news/microsofts-water-consumption-jumps-34-percent-amid-ai-boom/
Li, P., Yang, J., Islam, M.A., & Ren, S. (2025). Making AI less thirsty: Uncovering and addressing the secret water footprint of AI models. Communications of the ACM.https://dl.acm.org/doi/10.1145/3724499
Candisky, C. (2025). Ohio suspends data center tax breaks as costs soar past projections. The Columbus Dispatch.
The White House. (2026, March 4). Fact Sheet: The Ratepayer Protection Pledge.https://www.whitehouse.gov
Gold, R. & Puko, T. (2024, August). AI is sending data center energy demand soaring. The Wall Street Journal